A history-based parametric CAD sketch dataset with advanced engineering commands

Abstract

Modern computer-aided design (CAD) adopts history-based parametric modeling paradigm, in which the sketches play a crucial role to represent the design history, design intent, and design semantics of human engineers. Recent academic works include simple features, which are not suitable for the real-world engineering tasks. How to understand, generate and reconstruct CAD sketches with advanced engineering commands (primitives, operations and constraints) are still open challenges. To address these challenges, this paper is the first work to propose a history-based parametric CAD sketch dataset to support advanced engineering commands, named as HPSketch. Firstly, unlike existing simple datasets which are mainly composed of simple sketch primitives (typical line, arc, circle), HPSketch devises advanced primitives (such as parabola, hyperbola), operations (such as chamfer, fillet, rotation), and fruitful constraints. Secondly, HPSketch creatively propose a primitive selection command for advanced sketch operations, which can fully express the design intent and design knowledge for engineering design. This selection method is only available in our HPSketch dataset. Furthermore, unlike the simple fixed interactive constraints in previous sketch datasets, HPSketch presents a complicated, flexible constraint mechanism, which supports complicated engineering constraints. At present, HPSketch consists of 151,984 parametric sketches and 377,623 loops with 29 command types. The experiments show that the generated 2D sketches and 3D CAD models based on HPSketch are more complicated and more diverse than those based on existing sketch datasets. What is more encouraging is that sketches generated by HPSketch can be edited and re-designed by human engineers on mainstream industrial CAD softwares.